NETWORK EQUILIBRIUM APPROACHES TO URBAN TRANSPORT MARKETS: COMBINED MODELS AND EFFICIENT MATRIX ESTIMATION

A network equilibrium is a flow pattern in a given network that is consistent with supply and demand conditions on the transportation market and based on the behaviour of the travellers in the network. The travel behaviour assumed in this thesis includes user-optimal route choices and nested logit models for destination and mode choices. Trip distribution, modal split and trip assignment constitute the three stages of the combined models treated in this thesis. The Origin-Destination (OD) travel matrix estimation problem studies the reverse of the assignment problem: given traffic counts on certain road links of the network, estimate the flow of trips between zones. This travel matrix should reproduce the observed link flows and generate a reasonable OD matrix. The combined network equilibrium models are derived throughly. Two related estimation problems are investigated, analyzed and solved. Different techniques for the estimation procedure result in similar solutions and there are no empirical arguments for the theoretically better estimation formulation. In the Stockholm application strong arguments for a rejection of the traditional nested combined model are found. Of the different methods for OD matrix estimation using traffic counts, the third paper indicates that an approximate gradient based method suggested by Spiess is preferable. The OD matrices estimated by a method using second order information do in initial stages of the solution process seem to differ. This indicates that the OD matrix estimation problem, which is inherently non-convex, actually may lead to multiple solutions in the Stockholm applications. (A)